Performance of cardiovascular disease risk prediction equations in more than 14 000 survivors of cancer in New Zealand primary care: a validation study

Abstract

People with cancer have an increased risk of cardiovascular disease. Risk prediction equations developed in New Zealand accurately predict 5-year cardiovascular disease risk in a general primary care population in the country. We assessed the performance of these equations for survivors of cancer in New Zealand. For this validation study, patients aged 30-74 years from the PREDICT open cohort study, which was used to develop the New Zealand cardiovascular disease risk prediction equations, were included in the analysis if they had a primary diagnosis of invasive cancer at least 2 years before the date of the first cardiovascular disease risk assessment. The risk prediction equations are sex-specific and include the following predictors: age, ethnicity, socioeconomic deprivation index, family history of cardiovascular disease, smoking status, history of atrial fibrillation and diabetes, systolic blood pressure, total cholesterol to HDL cholesterol ratio, and preventive pharmacotherapy (blood-pressure-lowering, lipid-lowering, and antithrombotic drugs). Calibration was assessed by comparing the mean predicted 5-year cardiovascular disease risk, estimated using the risk prediction equations, with the observed risk across deciles of risk, for men and women, and according to the three clinical 5-year cardiovascular disease risk groups in New Zealand guidelines (<5%, 5% to <15%, and ≥15%). Discrimination was assessed by Harrell's C statistic. 14 263 patients were included in the study. The mean age was 61 years (SD 9) for men and 60 years (SD 8) for women, with a median follow-up of 5·8 years for men and 5·7 years for women. The observed cardiovascular disease risk was underpredicted by a maximum of 2·5% in male and 3·2% in female decile groups. When patients were grouped according to clinical risk groups, observed cardiovascular disease risk was underpredicted by less than 2% in the lower risk groups and overpredicted by 2·2% for men and 3·3% for women in the highest risk group. Harrell's C statistics were 0·67 (SE 0·01) for men and 0·73 (0·01) for women. The New Zealand cardiovascular disease risk prediction equations reasonably predicted the observed 5-year cardiovascular disease risk in survivors of cancer in the country, in whom risk prediction was considered clinically appropriate. Prediction could be improved by adding cancer-specific variables and considering competing risks. Our findings suggest that the equations are reasonable clinical tools for use in survivors of cancer in New Zealand. Auckland Medical Research Foundation, Health Research Council of New Zealand.